Skip to content
Get started

Resampler Microphone

The resampler microphone platform converts the sample rate of the audio from another microphone component. Components that need a specific sample rate, such as a voice assistant expecting 16 kHz audio, can then use a microphone that only supports higher sample rates.

Only the sample rate changes. The bits per sample and channels are selected with the microphone source settings of the source microphone. If the source microphone already delivers the target sample rate, the audio is passed through unchanged.

This platform only works on ESP32 based chips.

WARNING

Audio and voice components consume a significant amount of resources (RAM and CPU) on the device.

Crashes are likely to occur if you include too many additional components in your device’s configuration. In particular, Bluetooth/BLE components are known to cause issues when used in combination with Voice Assistant and/or other audio components.

# Example configuration entry
microphone:
  - platform: resampler
    microphone: source_microphone_id

Configuration Variables

Section titled “Configuration Variables”
  • microphone (Required, Microphone Source): The source microphone to resample. Supports 16 to 32 bits per sample and 1 or 2 channels.
  • sample_rate (Optional, positive integer): Sample rate to convert to. Must be between 8000 and 48000. Defaults to 16000.
  • filters (Optional, positive integer): The number of windowed sinc interpolation filters to use. Must be between 2 and 1024. Defaults to 16.
  • taps (Optional, positive integer): The number of taps per windowed sinc interpolation filter. Must be between 16 and 128 and divisible by 4. Defaults to 16.
  • All other options from Microphone Component.

Improving Quality

Section titled “Improving Quality”

Resampling is processor intensive and should be avoided where possible, for example by configuring the source microphone for the required sample rate directly. The audio quality is affected by the number of filters and the number of taps. Increasing the number of filters increases the memory load. Increasing the number of taps increases the CPU load.

When downsampling, frequencies above the new Nyquist limit are filtered out to avoid aliasing.

Example

Section titled “Example”

An I²S microphone reading 48 kHz stereo audio, resampled to 16 kHz mono with 16 bits per sample:

microphone:
  - platform: i2s_audio
    id: i2s_mic
    i2s_din_pin: GPIOXX
    adc_type: external
    sample_rate: 48000
    bits_per_sample: 32bit
    channel: stereo
  - platform: resampler
    id: resampled_mic
    microphone:
      microphone: i2s_mic
      channels: 0
      bits_per_sample: 16
    sample_rate: 16000

See Also

Section titled “See Also”